• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 21 subscribers
  • Views 1858 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMX2571EVM: Demodulated Signal shows interferences of unknown source

Alex Hoesl
Prodigy 180 points
Part Number: LMX2571EVM
Other Parts Discussed in Thread: LMX2571,

Hi there,

I am facing a problem that makes we wonder. I don't know where it comes from and how to get rid of it.

I'm stuck and maybe another point of view and input of others (you :-) ) could help to solve this problem!

I'm using the LMX2571 for 4-FSK modulation (GFSK via oversampling, I2S mode). I wanted to optimize PLL settings to get a better eye diagram. The symbol change has (i guess) an overshoot which I wanted to get rid of. But at the moment the demodulated signal is to much disturbed, I can't even control an eye-diagram. When I started to evaluate it was much better. I also was able to produce a sinusoidal FM modulation.

Looking now on a random GFSK signal it looks just like noise. That's why I'm showing you a change from FSK symbol 00 to 11 with oversampling, should be sinusoidal. But at the moment...well, take a look (markers are just somewhere, don't mean anything): 

Averaging the signal, it is sinusoidal.

The following screenshot was taken in the past, much better but already a little bit disturbed. I think it was even better, but I don't have a screenshot of this "far away past". 

I thought about corrupted I2S data, controlling huge arrays of data send to the LMX, but it looks all fine. It is repeating periodically, constant output. But it looks like the LMX sends the wrong deviations from time to time. Using a slower I2S data rate, increasing the baud rate, the disturbance get's even bigger, looking at the demodulated signal. But maybe it just stays equal high, looking only on the time domain. (markers show one periode, changing from symbol 11, 00, 11)

I played with the deviation, using lower deviation levels. 

Here a screenshot of the I2S signal (red: CLK; blue: FS; yellow: DATA). 

Detail of FS. If I had measured a bit better, it would have been more rectangular...

I also thought about wrong LMX settings, because I played with all kind of settings. But a can't find an fault or another setting to improve it.

Another thing it did, making the demodulation with another system, but looks similar. On other transmit frequencies the problem remains.

And I also thought about a OSC problem. I'm using the onboard LMX2571EVM oscillator at the moment. I tried with another one, doesn't improve it in any way.

What I didn't change yet: Replacing the PLL filter components.

I'm running out of options. Is my LMX maybe defective?

Any other ideas?

Thanks and kind regards,

Alex

  • 0
    Prodigy 180 points
    addition:
    I looked at the RF-signal one more time on spectrum analyzer. So without demodulation. When I was outputting a defined deviation in the past, so just one single frequency, RF-signal was showing a stable peak at desired deviation.
    If I output a single deviation right now, the signal shows a spread spectrum with a peak at it's center. The peak gets in the range of the desired deviation but is shifted.
    I tried it with different positive and negative deviations.
  • 0 in reply to Alex Hoesl
    TI__Expert 5370 points
    Dear Alex,

    Thank you for your interest in LMX2571.

    I have assigned an engineer to this detailed question. He should able to help early next week. He is very familiar with the FSK mode of the device. thank you for your patience,

    regards, Simon.
  • 0 in reply to Simon Damphousse
    TI__Guru** 109690 points
    Hi Alex,

    You mentioned that you were able to produce FM modulation in the past, can you make it happen again now? Were you using the I2S mode to produce FM?
    If you were able to produce FM using I2S mode, I assume you know how to produce GFSK using I2S mode.
    So the problem becomes why GFSK does not work? Here are some questions to you:
    1. what is the GFSK symbol rate?
    2. what is the over-sample factor (e.g 10x, 20x, ...) or the actual sampling rate?
    3. what is the loop bandwidth?
    4. What is the max. deviation?
    Please also check if this application note can help you: www.ti.com/.../snaa309.pdf
  • 0 in reply to Noel Fung
    Prodigy 180 points

    Hi Noel,

    thank you for the hints, the snaa309 definetely will be helpful.

    I guess the problem is related to a human-error a few days ago. One of the systems I use for demodulation is a receiver with antenna power (8V, ~2,4W) on its RF in. I forgot one day to use a DC-Block between Receiver and LMX EVM RF out. This probably caused the problem. 

    Thanks to TI, I was able to order some LMX2571 samples.

    Important question: Any recommendations about changing the IC on the EVM board? I can see that the chip has ground connections on backside/ bottom of PCB. Maybe not easy to remove and replace without damaging the conducting paths?!

    We also will order a second EVM for our development. We then can compare the performance of the repaired board and work on parallel.

    But to answer also your questions:

    1. Symbol rate is ~150kbit/s

    2. 5-times oversampling

    3. Actual loop bandwidth ~80-150kHz; this is definitely not final, still with the default components C1 C2 R2. As soon as we have final values for Fpd and Kpd, wen can adjust components.

    4. Max Dev. 45kHz.

    Kind regards,

    Alex

  • 0 in reply to Alex Hoesl
    TI__Guru** 109690 points
    Hi Alex,

    If you have some fancy soldering workstations, check the following apps note on how to rework. Key point is you have to heat both sides of the board.
    www.ti.com/.../snoa807.pdf
    www.ti.com/.../slva439a.pdf

    Your sampling rate is 750kHz, 150kHz loop bandwidth is not enough, you will get distortion. The loop bandwidth should be at least compatible with the sampling rate.
  • 0 in reply to Noel Fung
    Prodigy 180 points
    Hi Noel,

    I now experienced that our service technician solders stuff like that by hand... :-)

    After exchanging the IC, the problem still existed. But I guess I found the real problem.
    I thought about corrupted I2S data all the time. It was pure coincidence that I saw the FS cable solder joint was nearly broken. So the LMX sometimes got the signal, but most time the input data was useless....

    Topic closed, thanks Noel. Now I can have a close look on Loop Bandwidth.